Optimized neural network based path planning for searching indoor pollution source

Dehu Xiao, Yong Wang, Zhuo Cheng, Tianye Huang, Jun Yan

Abstract: Traditional path planning methods, such as A* and probabilistic roadmap, are seriously limited by the resolution and the number of samples that cannot meet the needs of complex indoor environments. Therefore, artificial neural network-based methods have become the mainstream due to the strong learning ability and robustness. In this paper, the neural network is used to control the agent according to the concentration and distance information, and a novel approach called adaptive neural evolution of augmenting topologies is proposed to optimize the neural network to improve its performance. Besides, the objective function is constructed by combining the pollution concentration and the residual energy of the agent to save energy and search time. Experiments show that our approach successfully finds an optimal path to pollution sources in different indoor environments while achieving energy saving and obstacle avoidance.